Abdulmajeed Alaskar

6 MONTHS vs. 20 YEARS. That’s the economic tension that caught my attention while reading Anthropic’s S-1. Over 80% of Anthropic’s Q3 revenue came from models released within the previous six months. Yet some of its infrastructure agreements extend for 20 years. Think about that. The technology generating today’s revenue evolves in months, while some of the financial commitments required to deliver it will last for decades. And the scale makes this extraordinary. 2023 revenue: $41M 2025 revenue: $4.6B Q2 2026 alone: $11.6B It even recorded $974M in operating profit in Q2. That widely reported $80.7B first-half net loss? Around $72.4B came from noncash accounting adjustments to financing instruments, largely driven by Anthropic’s rising valuation. But the number that really caught my attention was $417 BILLION. That’s approximately how much Anthropic disclosed in future minimum payment obligations under compute and infrastructure agreements entered into or amended since the beginning of 2026. Including: • Google: at least $114.5B • Amazon: at least $110B Both involve minimum spending commitments regardless of actual usage. Meanwhile, 80% of Anthropic’s first-half revenue was consumption-based. Although some customers have minimum commitments, many can reduce spending relatively quickly. Revenue can fluctuate overnight. Major infrastructure obligations cannot. Now consider how quickly the economics are changing. Anthropic’s own Opus 5.5 reduced customers’ cost per task by approximately 40% compared with its predecessor, while preserving margins. Remarkable progress. But what happens when newer chips or more efficient models allow competitors to deliver comparable intelligence at significantly lower costs? A company could experience explosive demand and record revenue while its margins come under pressure from contracts negotiated under yesterday’s technological assumptions. But there’s an equally powerful argument on the other side. When intelligence gets cheaper, people find more reasons to use it. Previously uneconomical tasks become viable. Companies automate more workflows. Entirely new markets emerge. Anthropic’s approximately 900% net revenue retention illustrates just how dramatically spending can expand among existing customers. If demand grows faster than prices fall, securing massive computing capacity years in advance could become an extraordinary competitive advantage. And that’s what makes this so fascinating. The same commitments could become either a competitive moat or an expensive constraint. Their terms and flexibility differ, and they aren’t all debt or immediately payable. There’s another twist: Claude is increasingly helping develop future versions of Claude, a process Anthropic calls “closing the loop.” The technology is effectively accelerating its own development, potentially making today’s economic assumptions outdated even faster. To me, this is a much more interesting question than whether AI is a bubble. The demand is real. The productivity gains are real. And the economic opportunity could be enormous. But being right about the future of AI doesn’t guarantee being right about the economics of building it. My biggest takeaway is that the greatest risk may not be that AI fails to deliver on its promise. It may be that AI advances so quickly that the financial commitments made to build its future struggle to keep up. Which raises the trillion-dollar question: Will these long-term infrastructure commitments become AI’s greatest competitive moat, or its most expensive anchor?

@MarioNawfal this is what happens when you throw economic figures together without understanding what they actually mean. Let’s dismantle this narrative with economics… 1. External debt is NOT sovereign debt. You’re comparing the external liabilities of an entire economy, including private-sector borrowers, against central bank reserves. These aren’t interchangeable balance-sheet items, and treating them as such is a fundamental analytical error. 2. Saudi Arabia is not running out of reserves. Official reserves stood at approximately $466 billion in September 2026, UP 3.4% year-on-year. You conveniently highlight the monthly decline while ignoring the annual comparison. 3. Your trade-collapse narrative ignores actual trade. Saudi Arabia recorded a SAR 155 billion merchandise trade surplus in the first half of 2026, up 53% year-on-year. 4. Your currency-crisis argument is a false choice. The IMF’s July 2026 assessment found Saudi Arabia’s reserves comfortable, projected approximately 14 months of import coverage, and reaffirmed that the dollar peg remains appropriate. 5. Your entire scenario assumes oil revenues effectively disappear. Yet Saudi crude shipments recovered to approximately 6.6 million barrels per day in September. Stress-testing an extreme scenario is legitimate. Presenting it as an approaching financial certainty is not. Nobody disputes that prolonged regional conflict creates economic risks. But risks are not insolvency, external debt is not a government bill due tomorrow, and managing reserves is not evidence of financial collapse. You have plenty of numbers, Mario. What you’re missing is the economics connecting them.